Conflict Management in Intelligent Robotic System based on FSM Approach


W. Jacak, K. Pröll, S. Dreiseitl - Conflict Management in Intelligent Robotic System based on FSM Approach - LECTURE NOTES IN COMPUTER SCIENCE, Vol. 2178, No. 2178, 2001, pp. 52-66


A robotics system consists of a community of independently acting robotics agents. Each of the robotics agents is under control of an intelligent software subagent. The group of the agents create the first level of the multiagent system i.e. the execution level. The main goal of a multiagent system is to solve a common task, which is split up into several subtasks that are distributed by a scheduling process (called Contract Manager) to the individual robotics agents in the system. Task distribution on the one side and task performance on the other side require two different units (called Conflict Manager) of cooperation and negotiation in the multiagent system. These two units are the main parts of the coordination level. The Contract Manager considers task distribution when a new job enters the system. This Manager has to direct autonomous agents by specifying individual goals (subtasks) for each of them. This can be done in two different ways. The Contract Manager can distribute subtasks to agents in a hierarchical way by simply assigning a subtask to an individual agent, or he can offer subtasks as service requests to the whole agent community. This entails a bidding process between contract manager and robotics agents in a market-like style The second unit - Conflict Manager - indicates cooperation and negotiation while each agent performs its assigned subtask. As the individual behaviour of all robotics agents involved in task achievement cannot be predicted in advance, the goals of two or more robotics agents can be in direct conflict with one another and the achievement of the whole task is endangered. The conflict situation in robotic multiagent system occurs when the distance between current positions of two or more robots is smaller as their security zones. The conflict between goals of robots can occur when the final positions are close to one another. In order to resolve a conflict situation, a negotiation process among all conflict parties has to take place. The result of the negotiation should be a solution, which results in goal achievement for all involved agents. In this paper we focus on of cooperation and negotiation in a multiagent system and present solutions for detecting and resolving conflicts between two or more robotics agents. Depending on the having knowledge different models of agent actor and world can be used to perform the negotiation and coordination process among concurrent acting agents. In case only partial knowledge about surrounding environment we can apply the learning based neural network model of agent actor (robot) and sensor signals based triangle representation of world geometry and in case of full knowledge about geometry of the static world (polygon representation) we use the model of robot based on the finite state machine formalism. These two cases lead to two different methods of negotiation in the conflict situation. The fist negotiation method is called local oriented negotiation for concurrent acting agent and allows resolving conflict between two agents by using one of the conflict avoidance strategies such as non-symmetric compromise, symmetric compromise and pragmatic strategy. The second negotiation method is called the negotiation for concurrent acting agent with sequential one step delay}. In the paper we show how the finite state machine model of the agent behaviour and the modified graph search procedure can be applied to solve the conflict avoidance problem.